An approximate nerve theorem

Thumbnail

Event details

Date 23.02.2016
Hour 10:1511:30
Speaker Primoz Skraba
(Artificial Intelligence Laboratory, Jozej Stefan Institute, Ljubljana)
Location
CM113
Category Conferences - Seminars
The nerve theorem [Borsuk 48] states that the homotopy type of a sufficiently nice topological space is captured by the nerve of a good cover of that space. In the case of persistence, we rarely compute the persistence diagram of a filtration exactly, but rather an approximation of it. In this talk we introduce the notion of an epsilon-good cover and its application to computing persistence. Rather than require a good cover, one where all the elements of the cover and their finite intersections are contractible, we define the notion of an epsilon good cover - one where its all elements and finite intersections are homologically trivial modulo epsilon-persistent classes (i.e. each element can have a small amount of topological noise). We show an approximation result for the persistence diagram of a filtration of the nerve and the underlying space which depends on epsilon and the maximal dimension of the nerve and show that this bound is tight.

Practical information

  • Informed public
  • Free

Organizer

  • Magdalena Kedziorek

Event broadcasted in

Share